DOCSLIB.ORG

A Universal Global Measure of Univariate and Bivariate Data Utility for Anonymised Microdata S Kocar

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
A Universal Global Measure of Univariate and Bivariate Data Utility for Anonymised Microdata S Kocar A universal global measure of univariate and bivariate data utility for anonymised microdata S Kocar CSRM & SRC METHODS PAPER NO. 4/2018 Series note The ANU Centre for Social Research & Methods the Social Research Centre and the ANU. Its (CSRM) was established in 2015 to provide expertise includes quantitative, qualitative and national leadership in the study of Australian experimental research methodologies; public society. CSRM has a strategic focus on: opinion and behaviour measurement; survey • development of social research methods design; data collection and analysis; data archiving and management; and professional • analysis of social issues and policy education in social research methods. • training in social science methods • providing access to social scientific data. As with all CSRM publications, the views expressed in this Methods Paper are those of CSRM publications are produced to enable the authors and do not reflect any official CSRM widespread discussion and comment, and are position. available for free download from the CSRM website (http://csrm.cass.anu.edu.au/research/ Professor Matthew Gray publications). Director, ANU Centre for Social Research & Methods Research School of Social Sciences CSRM is located within the Research School of College of Arts & Social Sciences Social Sciences in the College of Arts & Social Sciences at the Australian National University The Australian National University (ANU). The centre is a joint initiative between August 2018 Methods Paper No. 4/2018 ISSN 2209-184X ISBN 978-1-925715-09-05 An electronic publication downloaded from http://csrm.cass.anu.edu.au/research/publications. For a complete list of CSRM working papers, see http://csrm.cass.anu.edu.au/research/publications/ working-papers. ANU Centre for Social Research & Methods Research School of Social Sciences The Australian National University A universal global measure of univariate and bivariate data utility for anonymised microdata S Kocar Sebastian Kocar is a PhD candidate at the ANU Centre for Social Research & Methods and a data archivist in the Australian Data Archive at the Australian National University. His research interests include survey methodology, with a particular focus on online data collection, statistical disclosure control and higher education development. Abstract This paper presents a new global data utility steps: assumptions for statistical tests, statistical measure, based on a benchmarking approach. significance of the association, direction of the Data utility measures assess the utility of association and strength of the association (size anonymised microdata by measuring changes effect). in distributions and their impact on bias, variance and other statistics derived from the Since the model should be executed data. Most existing data utility measures have automatically with statistical software code or significant shortcomings – that is, they are a package, our aim was to allow all steps to limited to continuous variables, to univariate be done with no additional user input. For this utility assessment, or to local information loss reason, we propose approaches to automatically measurements. Several solutions are presented establish the direction of the association between in the proposed global data utility model. It two variables using test-reported standardised combines univariate and bivariate data utility residuals and sums of squares between groups. measures, which calculate information loss Although the model is a global data utility using various statistical tests and association model, individual local univariate and bivariate measures, such as two-sample Kolmogorov– utility can still be assessed for different types Smirnov test, chi-squared test (Cramer’s V), of variables, as well as for both normal and ANOVA F test (eta squared), Kruskal-Wallis H test non-normal distributions. The next important (epsilon squared), Spearman coefficient (rho) and step in global data utility assessment would Pearson correlation coefficient r( ). The model be to develop either program code or an R is universal, since it also includes new local statistical software package for measuring data utility measures for global recoding and variable utility, and to establish the relationship between removal data reduction approaches, and it can univariate, bivariate and multivariate data utility of be used for data protected with all common anonymised data. masking methods and techniques, from data reduction and data perturbation to generation Keywords: statistical disclosure control, data of synthetic data and sampling. At the bivariate utility, information loss, distribution estimation, level, the model includes all required data analysis bivariate analysis, effect size Methods Paper NO. 4/2018 iii Acronyms ANU Australian National University CSRM Centre for Social Research & Methods IL information loss K–S Kolmogorov–Smirnov SDC statistical disclosure control iv CENTRE FOR SOCIAL RESEARCH & METHODS AND SOCIAL RESEARCH CENTRE Contents Series note ii Abstract iii Acronyms iv 1 Introduction 1 2 Balancing disclosure risk and data utility 3 2.1 Disclosure risk measures 3 2.2 Information loss measures 4 3 A new user-centred global data utility measure 6 3.1 Structure of the utility model 6 3.2 Local univariate data utility after perturbation, suppression or removal 7 3.3 Local univariate data utility after global recoding 9 3.4 Local bivariate data utility 10 3.5 Data utility of data with removed variables 14 3.6 Multivariate data utility 14 4 Discussion 15 References 16 Tables and figures Figure 1 Exponential distribution of information loss, the function of local data utility 8 Table 1 Bivariate tests and statistics used for local bivariate data utility calculation 10 Figure 2 Local data utility calculation procedure 11 M ETHODS PAPER NO. 4/2018 v vi CENTRE FOR SOCIAL RESEARCH & METHODS AND SOCIAL RESEARCH CENTRE 1 Introduction Demand has been increasing for governments to disclosure, attribute disclosure, inferential release unit record files in either publicly available disclosure, and disclosure of confidential anonymised form or restricted-access moderately information about a population or model. SDC is protected form. Unit record files – which are primarily focused on identity disclosure – that is, basically tables that contain unaggregated identification of a respondent (Duncan & Lambert information about individuals, enterprises, 1989:207–208). On the one hand, releasing no organisations, and so on – are progressively data has zero disclosure risk with no utility; on being disseminated by government agencies the other, releasing all collected data including for both public use (i.e. anybody can access the identifying information offers high utility, but data) and research use (i.e. only researchers with the highest risk (Larsen & Huckett 2010). can access the data). These disseminated data Most research in the field has focused on the files can include detailed information – such privacy-constrained anonymisation problem – as medical, voter registration, census and that is, lowering information loss (IL) at certain customer information – which could be used for k-anonymity or l-diversity values, which is called the allocation of public funds, medical research the direct anonymisation problem. Since this and trend analysis (Machanavajjhala et al. 2007). problem can lead to significant IL, data could Publishing data about individuals allows quality be useless for specific applications, and the IL data analysis, which is essential to support an could be unacceptable to users (Ghinita et al. increasing number of real-world applications. 2009). Data producers should also be concerned However, it may also lead to privacy breaches about data users who would publish an analysis (Loukides & Gkoulalas-Divanis 2012:9777). Some based on statistics in anonymised files that are data are made publicly available, and some data not similar to corresponding statistics in the are offered in less safe settings (e.g. to registered original data. In that case, the data producers researchers on DVDs; see ‘five safes framework’ might have to correct any erroneous conclusions in Desai et al. [2016]). For instance, statistical that are reached, and such efforts might require organisations release sample microdata under greater resources than those needed to produce different modes of access and with different data with higher utility (Winkler 1998). The goal is, levels of protection. Data may be offered as therefore, to provide data products with reliable perturbed public use files, unperturbed statistical utility and controlled disclosure risk (Duncan & data to be accessed in on-site data labs, or even Stokes 2004, Templ et al. 2014). microdata under contract (Shlomo 2010) – also known as scientific use files – with a moderate The development of disclosure risk measures level of statistical disclosure control (SDC). has been ahead of the development of data utility This means that distributors should focus on measures, mostly as a result of the focus and data protection. Additionally, unit record files needs of data disseminators as data protectors, should be more restrictively protected in case especially when it comes to global measures of they are sensitive. Some variables in datasets those two data protection aspects. Research in can be sensitive as a result of the nature of the the field mostly focuses on developing unbiased information they contain – for example, data on approaches to measuring risk of identification,
Load more

Recommended publications
  • Chapter 9 Descriptive Statistics for Bivariate Data
    9.1 Introduction 215 Chapter 9 Descriptive Statistics for Bivariate Data 9.1 Introduction We discussed univariate data description (methods used to explore the distribution of the values of a single variable) in Chapters 2 and 3. In this chapter we will consider bivariate data description. That is, we will discuss descriptive methods used to explore the joint distribution of the pairs of values of a pair of variables. The joint distribution of a pair of variables is the way in which the pairs of possible values of these variables are distributed among the units in the group of interest. When we measure or observe pairs of values for a pair of variables, we want to know how the two variables behave together (the joint distribution of the two variables), as well as how each variable behaves individually (the marginal distributions of each variable). In this chapter we will restrict our attention to bivariate data description for two quan- titative variables. We will make a distinction between two types of variables. A response variable is a variable that measures the response of a unit to natural or experimental stimuli. A response variable provides us with the measurement or observation that quan- ti¯es a relevant characteristic of a unit. An explanatory variable is a variable that can be used to explain, in whole or in part, how a unit responds to natural or experimental stimuli. This terminology is clearest in the context of an experimental study. Consider an experiment where a unit is subjected to a treatment (a speci¯c combination of conditions) and the response of the unit to the treatment is recorded.
    [Show full text]
  • Visualizing Data Descriptive Statistics I
    Introductory Statistics Lectures Visualizing Data Descriptive Statistics I Anthony Tanbakuchi Department of Mathematics Pima Community College Redistribution of this material is prohibited without written permission of the author © 2009 (Compile date: Tue May 19 14:48:11 2009) Contents 1 Visualizing Data1 Dot plots ........ 7 1.1 Frequency distribu- Stem and leaf plots... 7 tions: univariate data.2 1.3 Visualizing univariate Standard frequency dis- qualitative data..... 8 tributions.... 3 Bar plot......... 8 Relative frequency dis- Pareto chart....... 9 tributions.... 4 Cumulative frequency 1.4 Visualizing bivariate distributions.. 5 quantitative data . 11 1.2 Visualizing univariate Scatter plots . 11 quantitative data.... 5 Time series plots . 12 Histograms....... 5 1.5 Summary . 13 1 Visualizing Data The Six Steps in Statistics 1. Information is needed about a parameter, a relationship, or a hypothesis needs to be tested. 2. A study or experiment is designed. 3. Data is collected. 4. The data is described and explored. 5. The parameter is estimated, the relationship is modeled, the hypothesis is tested. 6. Conclusions are made. 1 2 of 14 1.1 Frequency distributions: univariate data Descriptive statistics Descriptive statistics. Definition 1.1 summarize and describe collected data (sample). Contrast with inferential statistics which makes inferences and generaliza- tions about populations. Useful characteristics for summarizing data Example 1 (Heights of students in class). If we are interested in heights of students and we measure the heights of students in this class, how can we summarize the data in useful ways? What to look for in data shape are most of the data all around the same value, or are they spread out evenly between a range of values? How is the data distributed? center what is the average value of the data? variation how much do the values vary from the center? outliers are any values significantly different from the rest? Class height data Load the class data, then look at what variables are defined.
    [Show full text]
  • Rainbow Plots, Bagplots and Boxplots for Functional Data
    Rainbow plots, bagplots and boxplots for functional data Rob J Hyndman and Han Lin Shang Department of Econometrics and Business Statistics, Monash University, Clayton, Australia June 5, 2009 Abstract We propose new tools for visualizing large amounts of functional data in the form of smooth curves. The proposed tools include functional versions of the bagplot and boxplot, and make use of the first two robust principal component scores, Tukey's data depth and highest density regions. By-products of our graphical displays are outlier detection methods for functional data. We compare these new outlier detection methods with existing methods for detecting outliers in functional data, and show that our methods are better able to identify the outliers. An R-package containing computer code and data sets is available in the online supple- ments. Keywords: Highest density regions, Robust principal component analysis, Kernel density estima- tion, Outlier detection, Tukey's halfspace depth. 1 Introduction Functional data are becoming increasingly common in a wide range of fields, and there is a need to develop new statistical tools for analyzing such data. In this paper, we are interested in visualizing data comprising smooth curves (e.g., Ramsay & Silverman, 2005; Locantore et al., 1999). Such functional data may be age-specific mortality or fertility rates (Hyndman & Ullah, 2007), term- structured yield curves (Kargin & Onatski, 2008), spectrometry data (Reiss & Ogden, 2007), or one of the many applications described by Ramsay & Silverman (2002). Ramsay & Silverman (2005) and Ferraty & Vieu (2006) provide detailed surveys of the many parametric and nonparametric techniques for analyzing functional data. 1 Visualization methods help in the discovery of characteristics that might not have been ap- parent using mathematical models and summary statistics; and yet this area of research has not received much attention in the functional data analysis literature to date.
    [Show full text]
  • The Aplpack Package
    The aplpack Package September 3, 2007 Type Package Title Another Plot PACKage: stem.leaf, bagplot, faces, spin3R, ... Version 1.1.1 Date 2007-09-03 Author Peter Wolf, Uni Bielefeld Maintainer Peter Wolf <[email protected]> Depends R (>= 2.4.0), tcltk Description set of functions for drawing some special plots: stem.leaf plots a stem and leaf plot bagplot plots a bagplot faces plots chernoff faces spin3R for an inspection of a 3-dim point cloud License This package was written by Hans Peter Wolf. This software is licensed under the GPL (version 2 or newer) terms. It is free software and comes with ABSOLUTELY NO WARRANTY. URL http://www.wiwi.uni-bielefeld.de/~wolf/software/aplpack R topics documented: aplpack-package . 1 bagplot . 2 boxplot2D . 5 faces............................................. 6 spin3R . 8 stem.leaf . 9 Index 11 1 2 bagplot aplpack-package Another PLotting Package Description this package contains functions for plotting: chernoff faces, stem and leaf plots, bagplot etc. Details Package: aplpack Type: Package Version: 1.0 Date: 2006-04-04 License: GPL bagplot plots a bagplot faces plots chernoff faces scboxplot plot a boxplot into a scatterplot spin3R enables the inspection of a 3-dim point cloud stem.leaf plots a stem and leaf display Author(s) Peter Wolf Maintainer: Peter Wolf <[email protected]> References Examples bagplot bagplot, a bivariate boxplot Description compute.bagplot() computes an object describing a bagplot of a bivariate data set. plot.bagplot() plots a bagplot object. bagplot() computes and plots a bagplot. Usage bagplot(x, y, factor = 3, na.rm = FALSE, approx.limit = 300, show.outlier = TRUE, show.whiskers = TRUE, show.looppoints = TRUE, show.bagpoints = TRUE, show.loophull = TRUE, show.baghull = TRUE, create.plot = TRUE, add = FALSE, pch = 16, cex = 0.4, bagplot 3 dkmethod = 2, precision = 1, verbose = FALSE, debug.plots = "no", col.loophull="#aaccff", col.looppoints="#3355ff", col.baghull="#7799ff", col.bagpoints="#000088", transparency=FALSE, ..
    [Show full text]
  • Bivariate Descriptive Statistics: Introduction*
    OpenStax-CNX module: m46177 1 Bivariate Descriptive Statistics: Introduction* Irene Mary Duranczyk Suzanne Loch Janet Stottlemyer This work is produced by OpenStax-CNX and licensed under the Creative Commons Attribution License 3.0 Abstract This module included categorical-categorical, cateogorical-measurement, and measurement-measurement descriptive statistics It also contains information from the descriptive statistics chapter part of Collabo- rative Statistics collection (col10522) by Barbara Illowsky and Susan Dean. 1 Descriptive Statistics for Bivariate Data: Introduction 1.1 Student Learning Outcomes By the end of this chapter, the student should be able to: • Identify dierent types of relationships between variables: categorical-categorical, categorical-numerical, and numerical-numerical Categorical - Categorical • Know how to make and summarize a contingency table between two categorical variables • Discuss and describe patterns in contingency tables and supporting graphic summaries Categorical - Numerical • Be able to build side-by-side histograms, back-to-back stemplots, and multiple box-plots between a categorical and a numeric variable. • Compare and contrast multiple groups based on their shape, center, and spread. Numerical - Numerical • Discuss basic ideas of the relationship between two numeric variables including scatter plots, linear regression and correlation • Create and interpret a line of best t • Calculate and interpret the correlation coecient *Version 1.2: May 2, 2013 3:47 pm -0500 http://creativecommons.org/licenses/by/3.0/ http://cnx.org/content/m46177/1.2/ OpenStax-CNX module: m46177 2 1.2 Introduction In the previous chapters you have explored how to organize, summarize, and discuss both univariate cat- egorical and numerical variables. Usually more interesting questions are to examine relationships between sets of variables.
    [Show full text]
  • Package 'Aplpack'
    Package ‘aplpack’ July 29, 2019 Title Another Plot Package: 'Bagplots', 'Iconplots', 'Summaryplots', Slider Functions and Others Version 1.3.3 Date 2019-07-26 Author Hans Peter Wolf [aut, cre] Maintainer Hans Peter Wolf <[email protected]> Depends R (>= 3.0.0) Suggests tkrplot, jpeg, png, splines, utils, tcltk Description Some functions for drawing some special plots: The function 'bagplot' plots a bagplot, 'faces' plots chernoff faces, 'iconplot' plots a representation of a frequency table or a data matrix, 'plothulls' plots hulls of a bivariate data set, 'plotsummary' plots a graphical summary of a data set, 'puticon' adds icons to a plot, 'skyline.hist' combines several histograms of a one dimensional data set in one plot, 'slider' functions supports some interactive graphics, 'spin3R' helps an inspection of a 3-dim point cloud, 'stem.leaf' plots a stem and leaf plot, 'stem.leaf.backback' plots back-to-back versions of stem and leaf plot. License GPL (>= 2) URL http://www.wiwi.uni-bielefeld.de/lehrbereiche/statoekoinf/comet/ wolf/wolf\_aplpack NeedsCompilation no Repository CRAN Date/Publication 2019-07-29 07:50:12 UTC R topics documented: bagplot . 2 bagplot.pairs . 5 1 2 bagplot boxplot2D . 6 faces............................................. 8 hdepth . 10 iconplot . 11 plothulls . 26 plotsummary . 28 puticon . 29 skyline.hist . 34 slider . 37 slider.bootstrap.lm.plot . 41 slider.brush . 42 slider.hist . 44 slider.lowess.plot . 45 slider.smooth.plot.ts . 46 slider.split.plot.ts . 47 slider.stem.leaf . 48 slider.zoom.plot.ts . 49 spin3R . 50 stem.leaf . 51 bagplot bagplot, a bivariate boxplot Description compute.bagplot() computes an object describing a bagplot of a bivariate data set.
    [Show full text]
  • Further Maths Bivariate Data Summary
    Further Maths Bivariate Data Summary Further Mathematics Bivariate Summary Representing Bivariate Data Back to Back Stem Plot A back to back stem plot is used to display bivariate data, involving a numerical variable and a categorical variable with two categories. Example The data can be displayed as a back to back stem plot. From the distribution we see that the Labor distribution is symmetric and therefore the mean and the median are very close, whereas the Liberal distribution is negatively skewed. Since the Liberal distribution is skewed the median is a better indicator of the centre of the distribution than the mean. It can be seen that the Liberal party volunteers handed out many more “how to vote” cards than the Labor party volunteers. Parallel Boxplots When we want to display a relationship between a numerical variable and a categorical variable with two or more categories, parallel boxplots can be used. Example The 5-number summary of each class is determined. Page 1 of 19 Further Maths Bivariate Data Summary Four boxplots are drawn. Notice that one scale is used. Based on the medians, 7B did best (median 77.5), followed by 7C (median 69.5), then 7D (median 65) and finally 7A (median 61.5). Two-way Frequency Tables and Segmented Bar Charts When we are examining the relationship between two categorical variables, the two-way frequency table can be used. Example 67 primary and 47 secondary school students were asked about their attitude to the number of school holidays which should be given. They were asked whether there should be fewer, the same number or more school holidays.
    [Show full text]
  • BIVARIATE DATA 7.1 Scatter Plots
    Page 1 of 7 BIVARIATE DATA 7.1 Scatter plots: A scatter plot, scatterplot, or scattergraph is a type of mathematical diagram using Cartesian coordinates to display values for two variables for a set of data. The data is displayed as a collection of points, each having the value of one variable determining the position on the horizontal axis and the value of the other variable determining the position on the vertical axis. This kind of plot is also called a scatter chart, scatter gram, scatter diagram, or scatter graph. Overview[ A scatter plot is used when a variable exists that is below the control of the experimenter. If a parameter exists that is systematically incremented and/or decremented by the other, it is called the control parameter or independent variable and is customarily plotted along the horizontal axis. The measured or dependent variable is customarily plotted along the vertical axis. If no dependent variable exists, either type of variable can be plotted on either axis and a scatter plot will illustrate only the degree of correlation (not causation) between two variables. A scatter plot can suggest various kinds of correlations between variables with a certain confidence interval. For example, weight and height, weight would be on x axis and height would be on the y axis. Correlations may be positive (rising), negative (falling), or null (uncorrelated). If the pattern of dots slopes from lower left to upper right, it suggests a positive correlation between the variables being studied. If the pattern of dots slopes from upper left to lower right, it suggests a negative correlation.
    [Show full text]
  • Package 'Aplpack'
    Package ‘aplpack’ February 15, 2013 Type Package Title Another Plot PACKage: stem.leaf, bagplot, faces, spin3R, and some slider functions Version 1.2.7 Date 2012-10-03 Author Peter Wolf, Uni Bielefeld Maintainer Peter Wolf <[email protected]> Depends R (>= 2.8.0), tcltk Suggests tkrplot Description set of functions for drawing some special plots: stem.leaf plots a stem and leaf plot stem.leaf.backback plots back-to-back versions of stem and leafs bagplot plots a bagplot faces plots chernoff faces spin3R for an inspection of a 3-dim point cloud slider functions for interactive graphics License GPL (>= 2) URL http://www.wiwi.uni-bielefeld.de/com/wolf/software/aplpack.html Repository CRAN Date/Publication 2012-12-11 07:55:36 NeedsCompilation no R topics documented: bagplot . .2 bagplot.pairs . .5 boxplot2D . .6 faces.............................................8 hdepth . 10 plotsummary . 11 1 2 bagplot skyline.hist . 13 slider . 15 slider.bootstrap.lm.plot . 20 slider.brush . 21 slider.hist . 22 slider.lowess.plot . 24 slider.smooth.plot.ts . 25 slider.split.plot.ts . 26 slider.stem.leaf . 27 slider.zoom.plot.ts . 28 spin3R . 29 stem.leaf . 30 Index 32 bagplot bagplot, a bivariate boxplot Description compute.bagplot() computes an object describing a bagplot of a bivariate data set. plot.bagplot() plots a bagplot object. bagplot() computes and plots a bagplot. Usage bagplot(x, y, factor = 3, na.rm = FALSE, approx.limit = 300, show.outlier = TRUE, show.whiskers = TRUE, show.looppoints = TRUE, show.bagpoints = TRUE, show.loophull = TRUE, show.baghull = TRUE, create.plot = TRUE, add = FALSE, pch = 16, cex = 0.4, dkmethod = 2, precision = 1, verbose = FALSE, debug.plots = "no", col.loophull="#aaccff", col.looppoints="#3355ff", col.baghull="#7799ff", col.bagpoints="#000088", transparency=FALSE, ..
    [Show full text]
  • Investigate Bivariate Measurement Data Written by Jake Wills – Mathsnz – [email protected]
    NCEA Level 3 - 3.9 Bivariate Data students.mathsnz.com Level 3 – AS91581 – 4 Credits – Internal Investigate Bivariate Measurement Data Written by Jake Wills – MathsNZ – [email protected] Achievement Achievement with Merit Achievement with Excellence Investigate bivariate Investigate bivariate Investigate bivariate measurement data. measurement data, with measurement data, with justification. statistical insight. Part 1: Problem ................................................................................................................................................. 2 Part 2: Plan ........................................................................................................................................................ 4 Part 2.1: Identifying the Variables .............................................................................................................. 4 Part 2.2: Naming the Source ...................................................................................................................... 6 Part 3: Data – Using NZGrapher ..................................................................................................................... 8 Part 4: Analysis .................................................................................................................................................. 9 Part 4.1: Trend ............................................................................................................................................... 9 Part 4.2: Association..................................................................................................................................
    [Show full text]
  • Statistics Statistics- the Branch of Mathematics That Deals with The
    Statistics Statistics- the branch of mathematics that deals with the collection, organization, analysis, and interpretation of numerical data. Dot Plot- a statistical chart consisting of data points plotted on a fairly simple scale. Histogram- a diagram consisting of rectangles whose area is proportional to the frequency of a variable and whose width is equal to the class interval. Box Plot- a graphic way to display the median, quartiles, maximum and minimum values of a data set on a number line. Interquartile Range- the range of values of a frequency distribution between the first and third quartiles. Standard Deviation- a measure of how spread out numbers are. Outlier- A value that "lies outside" (is much smaller or larger than) most of the other values in a set of data. y < Q1 − 1.5 × IQR or y > Q3 + 1.5 × IQR Frequency- the rate at which something occurs or is repeated over a particular period of time or in a given sample. Frequency Table- a table that shows the total for each category or group of data. Relative Frequency- how often something happens divided by all outcomes. Joint Frequency- It is called joint frequency because you are joining one variable from the row and one variable from the column. Joint Relative Frequency- the ratio of the number of observations of a joint frequency to the total number of observations in a frequency table. Marginal Frequency- the total in a row or column in a two way table. Marginal Relative Frequency- the ratio of the sum of the joint relative frequency in a row or column and the total number of data values.
    [Show full text]
  • Determination of Dependence Structure by Using Graphical Tools for Bivariate Continuous Data
    Çankaya Üniversitesi Fen-Edebiyat Fakültesi, Journal of Arts and Sciences Sayı: 12 / Aralık 2009 Determination of Dependence Structure by Using Graphical Tools for Bivariate Continuous Data Özlem Ege Oruç*, Zeynep F. Eren Doğu* Abstract The dependence between a pair of continuous variates can be numerous with potentially surprising implications. Global dependence measures, such as Pearson correlation coefficient, can not reflect the complex dependence structure of two variables. For bivariate data set the dependence structure can not only be measured globally, but the dependence structure can also be analyzed locally. Thus, scalar dependence measures are extended to local dependence measures. In this study, determination of local dependence structure of bivariate data is discussed. For this, some graphical methods called chi-plot and local dependence map are used and examples with simulated and economic data set are illustrated. Keywords: Local dependence, Correlation, Chi-plot, Local dependence map. Özet İki değişken arasındaki bağımlılık yapısı hakkında bilgi sahibi olmak oldukça önemlidir. Günümüzde yay- gın olarak kullanılan global bağımlılık ölçüleri (Örneğin: Pearson korelasyon katsayısı gibi), iki değişken arasın- daki karmaşık bağımlılık yapısını tam olarak yansıtamamaktadır. Bu nedenle iki değişken arasındaki bağımlılık yapısını sadece global bağımlılık ölçüleri ile değil lokal bağımlılık ölçüleri ile de incelemek gerekir. Çalışmada, iki değişken arasındaki bağımlılık yapısının belirlenmesinde kullanılan lokal bağımlılık ölçüleri incelenmiş ve bu yapının belirlenmesinde oldukça pratik araçlar olan ki-grafiği ve yerel bağımlılık haritaları kullanılarak uy- gulama yapılmış ve yorumlanmıştır. Anahtar Kelimeler: Lokal bağımlılık, Korelasyon, Ki-grafiği, Lokal bağımlılık haritası. * Dokuz Eylül University, Faculty of Arts and Sciences, Department of Statistics, Tinaztepe Campus 35160 Buca İzmir, TURKEY. e-mails: [email protected], [email protected] 83 Determination of Dependence Structure by Using Graphical Tools for Bivariate Continuous Data 1.
    [Show full text]